- Made the getBoardState method in Referee.java static.
- Created NeuralNetworkPlayer.java, though I don't know how to make the computer use it.
- Updated the Player interface to include passing the PlayerModel. Most of the current com agents ignore the data, but it is now available.
- Updated the train function in PlayerModel.java.
Default is AlphaBeta.
Updated ActionListener for UserChooserFrame so that selecting from the list (or typing a new name and pressing RETURN) also triggers the action.
Added the player and computer algorithm names to the title bar e.g. "Bob vs. Alpha-Beta".
- Integrated the ANN with the game. The network now predicts a user move, completely ignores it, and trains itself on the players actual move. This integration also included implementing two new functions. The first translates a board state to a boolean array to correspond with input nodes. The second translates a move to a boolean array to correspond with output nodes.
- Added support for the player model inside Referee.java; high scores should now persist over a single execution of the program.
- Refactored PlayerModel.java to support game logging. All games are now logged so that we can track overall progress.
- Added scaffolding to allow saving and importing of PlayerModel.java. It is not yet functional, but it will be with two function implementations and then the appropriate calls.
Fixed instantiation where a new MouseListener was used to each tile - now only one is used for all Tiles.
Fixed issue where the MouseListener was being added to each tile twice.
Renamed Player implementations and changed Player to interface, with the goal of using abstraction to make the human and computer interactions with the game identical for ease of simulation.
While not as clean as using ChangeListener, it completely decouples the Model and View aspects,
and also avoids the need to manually paint() a hierarchy of components (a task best handled by Swing/AWT).
